as usual, a lot of negative comments from HN on the approach (which is admittedly against the grain) but I've used it for a relatively large and successful web app (rails back end) and it has worked out well compared to the SPA architectures I've had to deal with
i think the philosophical roots in REST/HATEOAS are worth considering, if you are open minded:
Have you tried something like prmd [0]? And if so, what are your thoughts on it?
I used it at my previous job, but didn't find it gave me that many benefits, although there wasn't anything particularly wrong with it either. It just seems like a challenging problem to properly solve with any organization.
I was initially really interested in GraphQL and Relay, but when I tried em out I found em lacking for our small engineering team. I feel kinda similar with all these other alternative tools. I think if you have any API stability problems, you need to tackle it at an organizational level. The API team needs to communicate openly with the web frontend and mobile app team (which might include a cross-section of highly-mixed-specialty people) to make sure all changes are fully backwards compatible. I don't think the tools do a great job at helping you tackle complex issues.
I have used swagger a bit for JSON API development, but didn't find it added much over just keeping it simple and reacting to client needs.
I typically split my externally facing JSON API out from the main application (which uses HTML/intercooler) because the use cases are typically so different: the JSON API needs to be general and abstract whereas the web application UI has a lot of fiddly little specific data needs that are better implemented on the back end, where you have full access to the domain model and a proper query language.
Sometime ago I had an idea that SPAs exist mostly because animated transitions between pages make managers happy, not to serve any actual user need. If that's true, maybe we could add support for animated transitions between server-side rendered pages, and make SPAs mostly go away.
Then I realized we could do that easily with a JS library! We could give some divs a "sticky-id" attribute, so they stay as part of the DOM tree when the page reloads, only their contents get replaced with the contents of divs from the new page having the same "sticky-id". That way there's no flash to white, and CSS transitions on sticky elements get a chance to fire. The whole thing could be handled by a generic onclick handler on each link. And the links would support opening in new tab as well, the JS would only fire when they are clicked in the same tab. I think that would be the most conservative way to have an SPA-like experience with a minimum of JS. Unfortunately I was too lazy to implement it :-/
EDIT: It's interesting that most replies to this comment seem to talk about whole page transitions, like fading out the whole page. But as I envisioned it, the key part of the idea is that we can get element-by-element transitions without writing custom JS, just by adding "sticky-id" attribute to some elements and letting their CSS transitions do the work.
Turbolinks looks neat, but our Ruby app selectively includes CSS and Javascript snippets outside the <body> tag depending on what page is loaded (using :content_for declarations in our templates). I am guessing that Turbolinks will not work for us in this case, as the relevant external files will not be called on page load?
Neat! Though the point of my idea is having animated transitions between pages, by replacing contents of sticky elements and letting their CSS transitions work as usual. As far as I can tell, turbolinks doesn't do that.
Yeah, well... that still requires you to write custom JS. The thing I'm imagining would be based only on markup - which sticky ids exist in both pages. But the idea is really similar, thanks for the link!
It's independent of Rails. It simply hooks the all calls to <a href> tags and replaces them with ajax get requests to the url.
It works really nice with Rails because of how easy Rails makes it return a piece of javascript to the page (this is called SJR) so there is some magic included for making form submissions work which is part of the Rails turbolinks gem. The javascript that the server returns to the page essentially calls Turbolinks.visit("/url-form-submission-would-redirect-to-after-completion") so that the expected navigation takes place. It's quite clever. I think you would have to replicate that in the Java world to get the full experience. This SO answer alludes to how you might replicate SJR https://stackoverflow.com/questions/26750738/java-spring-ret...
So I think out of the box you can expect clicking links to work (and that alone provides a great speedup), but ajax form submissions will require some work. If you look at some articles on how form submissions with turbolinks make use of SJR you could likely figure it out and implement it. It's overall not a very complicated mechanism.
I suppose Rails .erb templating engine also has some helpers which the form submission side of turbolinks relies on. Namely it includes a data attribute data-remote and that's the thing turbolinks hooks into to prevent the default behavior of the form submission and and instead fire the ajax request. In theory so long as your forms include that it will work if you have the SJR stuff in place on your back-end.
I've done something not the same, but a compromise previously using the Window object's `beforeunload` event to fade out everything in the DOM smoothly, and then using the `beforeload` and `load` events to fade the new DOM back in. It's a bit of a halfway between having sliding transitions, etc. You might be able to simulate that even further than I ever tried.
The primary benefit I see from SPAs is to enable pages that changes in a wide variety of ways without having to reload the whole thing at any point.
Funny thing, you can actually serve each page as a separate, stand-alone JS "app/page" and still get this benefit (e.g. each page in your app becomes separate webpack entry point).
I tried this library before but Vue does pretty much the same thing and is way more powerful. Additionally, starting out with Vue makes transitioning to a SPA much easier when needed.
I also encountered some bugs when mixing the 2 during a transition phase that broke my requests. If I had to start over again I would only use Vue starting with the drop in JS library (no webpack/preprocessing)
It may be worth mentioning that Vue is an entire framework with a crapload of features and caveats and whatnot and thus not exactly an equivalent of an especially flexible PJAX library which is the best definition I can come up with when defining Intercooler
Vue is not a framework, at least not in it's simplest form - drop in UI library.
I agree that Vue does much more than Intercooler and in theory, it could be seen as an overkill. But in reality, a developer will probably use it in conjunction with jQuery and then he might be better off replacing these two with Vue.js
I mean, regardless of anecdotal specific usage, vue is a framework. And a nice one too! Just entirely too much for something that can be fulfilled by a simple pjax tool
Seems useful when you've written a traditional server-side rendered app and you just want to "sprinkle" some AJAX onto it.
Back when I used PHP 5-6 years ago, I would use an "AJAX" helper built into the CakePHP MVC framework which would generate JavaScript for really basic AJAX requests. It was great for rapid prototyping and even production for admin panels and such. It's cool to see a similar, client-sided, framework-agnostic library like this with similar functionality.
I'm missing something. What's the problem they are solving? It looks like they are offering a more limited way to make AJAX calls. Why?
To add -- it depends on jQuery? Yeah, I'm gonna have to pass. What this thing does is so small that I can't fathom why they would need to import all of jQuery just to provide a new way to do AJAX requests.
I'll just keep using React, it probably already solves whatever problem is trying to be solved here.
if it helps you can think of intercooler as HTML++, addressing the shortcomings of HTML (only supporting GET/POST, only allowing clicks/submissions to generate requests, only targeting full page replacements, etc.) rather than trying to be something like React.
It isn't right for everything, but it works well for me w/ a rails back end and there are some good theoretical reasons for using HTML if you like REST/HATEOAS. I linked a few articles up thread you can read if you are interested.
This is fine if you build a website or a small app. But I wouldn’t recommend this approach for a more serious/big app. You will end up with unmaintainable soup.
My experience building a large flask app with jinja2 templating (large in that the app is ultimately a mix of ERP/manufacturing management/customer+supplier portal). The intercooler approach combined with flask blueprints and template folders is extremely maintainable. It is easy to comprehend all aspects of what is going on at any time. Can get a decent developer up to speed to be productive with it, within days. The vast majority of businesses would never need a heavier framework than this.
Like another poster said. This would be useful if you’re using a server side rendering framework like ASP.NET MVC and you just needed a little bit of AJAX.
You would maintain most of your HTML, routing logic, etc. server side.
as usual, a lot of negative comments from HN on the approach (which is admittedly against the grain) but I've used it for a relatively large and successful web app (rails back end) and it has worked out well compared to the SPA architectures I've had to deal with
i think the philosophical roots in REST/HATEOAS are worth considering, if you are open minded:
http://intercoolerjs.org/2016/01/18/rescuing-rest.html
http://intercoolerjs.org/2016/05/08/hatoeas-is-for-humans.ht...